Methodological and regulatory considerations for causal AI in drug development

Area of Science:

• Drug development and regulatory science
• Causal inference methodologies
• Artificial intelligence applications

Background:

• Artificial intelligence (AI) presents significant opportunities for advancing drug development processes.
• Regulatory agencies are increasingly issuing guidance on AI adoption in pharmaceuticals.
• The application of AI to causal inference, crucial for understanding treatment effects and informing regulatory decisions, remains underdeveloped.

Purpose of the Study:

• To review current regulatory activities concerning AI in drug development.
• To examine statistical methodologies for AI-driven causal inference.
• To identify key regulatory challenges and demonstrate AI's value in causal inference across various data sources and study designs.

Main Methods:

• Review of regulatory agency documents and guidance on AI.
• Examination of statistical and machine learning techniques for causal inference.
• Case illustrations of AI application in diverse drug development data and studies.

Main Results:

• Limited current regulatory guidance specifically addresses AI for causal inference in drug development.
• Various AI methodologies show potential for enhancing causal inference from complex datasets.
• AI can add significant value by integrating diverse data sources to strengthen treatment effect estimation.

Conclusions:

• There is a need for clearer regulatory pathways for AI-driven causal inference in drug development.
• Further research and validation of AI statistical methods are required for regulatory acceptance.
• AI holds substantial promise for improving the accuracy and efficiency of causal inference in pharmaceutical research and regulatory decision-making.